API Reference

Complete API documentation for Perspt’s 7-crate workspace architecture.

digraph crates { rankdir=LR; node [shape=box, style="rounded,filled", fontname="Helvetica", fontsize=10]; cli [label="perspt-cli", fillcolor="#4ECDC4", href="perspt-cli.html"]; core [label="perspt-core", fillcolor="#45B7D1", href="perspt-core.html"]; tui [label="perspt-tui", fillcolor="#96CEB4", href="perspt-tui.html"]; agent [label="perspt-agent", fillcolor="#FFEAA7", href="perspt-agent.html"]; policy [label="perspt-policy", fillcolor="#DDA0DD", href="perspt-policy.html"]; sandbox [label="perspt-sandbox", fillcolor="#F8B739", href="perspt-sandbox.html"]; store [label="perspt-store", fillcolor="#87CEEB", href="perspt-store.html"]; }

Crate Overview

Crate APIs

Crate Summary

Crate

Description

Key Types

perspt-cli API

CLI entry point with 10 subcommands

Commands, Cli

perspt-core API

LLM provider, config, memory

GenAIProvider, Config

perspt-agent API

SRBN engine for autonomous coding

SRBNOrchestrator, TaskPlan, Energy

perspt-tui API

Terminal UI components

AgentApp, Dashboard, DiffViewer

perspt-policy API

Security policy engine

PolicyEngine, Sanitizer

perspt-sandbox API

Process isolation

SandboxedCommand

perspt-store API

DuckDB session persistence

SessionStore, LlmRequestRecord

Architecture Quick Reference

🖥️ perspt-cli

10 Subcommands: chat, agent, init, config, ledger, status, abort, resume, logs, simple-chat

perspt-cli API
🔌 perspt-core

Thread-safe LLM: GenAIProvider with Arc<RwLock>

perspt-core API
🤖 perspt-agent

SRBN Engine: Orchestrator, LSP, Tools, Ledger

perspt-agent API
🎨 perspt-tui

Ratatui UI: Dashboard, DiffViewer, ReviewModal

perspt-tui API
🛡️ perspt-policy

Security: Starlark rules, command sanitization

perspt-policy API
📦 perspt-sandbox

Isolation: Resource limits, process control

perspt-sandbox API
💾 perspt-store

Persistence: DuckDB sessions, LLM logging

perspt-store API

Common Patterns

Error Handling

All crates use anyhow::Result for error propagation:

use anyhow::{Context, Result};

fn example() -> Result<()> {
    do_something()
        .context("Failed to do something")?;
    Ok(())
}

Async Operations

Built on Tokio async runtime:

use tokio::sync::mpsc;

async fn stream_response(sender: mpsc::Sender<String>) -> Result<()> {
    // Stream tokens as they arrive
}

Thread-Safe Sharing

Use Arc for sharing across tasks:

use std::sync::Arc;

let provider = Arc::new(GenAIProvider::new()?);
let provider_clone = Arc::clone(&provider);

tokio::spawn(async move {
    provider_clone.generate_response(...).await
});

See Also